cancel
Showing results for 
Search instead for 
Did you mean: 

Haswell overclock

jfrech14
Level 7
Hey there all! I am playing around with my clocks and have gotten my 4770k to be stable through 10 iterations of intel burn test on maximum stress setting at 4.4 ghz with the ram at 2.4 ghz and it reached 83C I have a 360 rad and a 240 rad custom loop. Does this sound like a decent result? I'm going to see how stable I can get the CPU clock without a ram overclock but ibt came up with just over 128 gigaflops which seems to be a nice result. Just hotter than I hoped 😕 be back soon with another overclock. Let me know your thoughts please 🙂

I know it isn't really the best chip produced but if will do 4.6 with the ram at 2.4 just won't pass intel. It passes cinebench and realbench but once I touch 4.7 windows barely starts. So not great chip not terrible chip?
Asus Maximus VI Extreme
Intel i7 4770K
8 GB G.Skill Trident X 2600MHz
1TB Samsung Evo SSD
1TB Western Digital Caviar Black
Corsair HX1000W PSU
SLI - Custom Watercooled EVGA GTX 770 SC 4GB
XSPC Custom Watercooling loop AX360 dual pump with 360 Rad and 240 Rad
NZXT Switch 810
NZXT Hue LED lighting
272 Views
31 REPLIES 31

jab383
Level 13
Nice results for a days work and it sounds like you had some fun doing it. I think the temperatures you're seeing say that's as far as you can go with the CPU and cooling you have. I think my comments above hold -- delidding is the step you'd have to take for the next step of CPU speed.

Have you submitted that score and checked where you stand on the leaderboard? When you do, be sure to filter the standings to just Maximus to avoid comparing with 6-core quad RAM channel extreme editions.

Don't put too much effort into the OpenCL score. It's great for an Nvidia GPU, and about 20000 higher than mine. See the Realbench section of the forum for discussions on Nvidia's treatment of OpenCL. Another thread in that section discusses optimizing a system for benchmarks -- streamlining Windows, task priorities, etc. I think you'll get more out of that than from OpenCL.

The area I think needs most attention is RAM. I'm actually nervous about the RAM voltage you report. It has been said by Intel that RAM voltage over 1.65 is hazardous to Haswell health. With RAM rated for 2600MHz, I would expect it would need less than 1.65v when running at 2400MHz in the Realbench screenshot. My RAM is rated 2400 and runs at 2400 with 1.63volts with very tight timing. With relaxed timing, mine runs at 2400MHz with 1.56volts.

The ROG guide
http://rog.asus.com/242142013/labels/rog-exclusive/maximus-vi-series-uefi-guide-for-overclocking/
goes into RAM tuning and I learned much from it.

The screenshot says the core clock is 800MHz. That means speedstep (EIST) is on. You might get better performance, and consistency is more likely, with EIST disabled, turbo enabled and fully manual OC. The guide goes into that stuff, too.

Jeff

I am 7000 off of your score so no worries I'm taking your words to heart for sure 😉
Asus Maximus VI Extreme
Intel i7 4770K
8 GB G.Skill Trident X 2600MHz
1TB Samsung Evo SSD
1TB Western Digital Caviar Black
Corsair HX1000W PSU
SLI - Custom Watercooled EVGA GTX 770 SC 4GB
XSPC Custom Watercooling loop AX360 dual pump with 360 Rad and 240 Rad
NZXT Switch 810
NZXT Hue LED lighting

jfrech14
Level 7
OK guys... well I am done for now I made it to the first page of the 4 core realbench leaderboard with a score of 87000. I think I can be happy with that for a while since I never really planned on doing competitive overclocking when I made the computer. But the potential was there and it's extremely addictive. My next step is another 8 gb of ram and eventually another 770 or sell them and upgrade to 780s and I will do another binge on overclocking at that point probably but I have gone through most of that manual and done most of it and think I am just at the voltage/thermal limit of my chip. I got realbench to run stable at 4607 MHz ram at 2404 MHz 10-12-12-29-2T and even got ~2600 MHz 10-12-12-27-1T but that benchmark was lower than the first due to my GPUs not being overclocked by accident. At 4.4 ghz I got a stable 2600 MHz 7-11-8-22-1T which was cool but could only get a 40000 in men tweakit. I managed to overclock my GPUs +90 MHz core and +900 MHz memory so all in all... I am pretty happy with that result. That huge gpu overclock got me the exact same score as a much smaller overclock... Did I hit bandwidth limit for x8? Lol. But anyway I delidded my CPU last night so I am running my stable/everyday use 4.4 GHz overclock in all the benchmarks to make sure my temperatures are lower than before just as a paranoia check. Thank you guys for you help. It has been a ton of fun and will be happening again. Maybe when the new CPU comes out I will go ape-**** on this 4770k and get the new one since I heard it is supposed to be Z87 lga 1150 compatible? We will see. But you guys rock!

and Jeff... I bow to your 90000 🙂 I wanted that top 4770K spot but I just couldn't get those high frequencies stable. You have an awesome setup there!

All the best guys and see you again soon 🙂
Josh


Ps.. I forgot to really check temps before because I wasn't expecting to delid. But I did intel burn test at 4.4 ghz 2400 MHz ram on maximum stress level and core 1 got to 80c cores 0 and 2 got to 78 and core 3 was maybe around 73 as max temps. Does this sound like a decent delid job? I may get a better tim later on and use it but I used as5 because I had it and saw that it didn't do too badly in tests. Thanks again 🙂 I learned a ton on this adventure
Asus Maximus VI Extreme
Intel i7 4770K
8 GB G.Skill Trident X 2600MHz
1TB Samsung Evo SSD
1TB Western Digital Caviar Black
Corsair HX1000W PSU
SLI - Custom Watercooled EVGA GTX 770 SC 4GB
XSPC Custom Watercooling loop AX360 dual pump with 360 Rad and 240 Rad
NZXT Switch 810
NZXT Hue LED lighting

jab383
Level 13
AS5 is as good as TIM gets and still be electrically non-conductive. It's a big and hazardous step to get to a better TIM -- liquid metal. That stuff is electrically conductive and takes special care, especially under the lid.

I understand the addiction. I got it unexpectedly too. I'm already at the point where I don't respect my CPU enough to preserve it. This dog could turn to toast and I would gladly get another. The reason I don't anyway is the Haswell lottery - as much chance of getting one that's worse as one that's better - and saving my pennies for the impending Devils Canyon.

My CPU clock isn't all that high - x47 multiplier. I ran up against the CPU's limits with x47, Bclk 1% high and 1.565 volts on the cores. That's just stable enough to do the benchmark and scored under 90000. Memory tweaking did the rest. You have a great start and the 2600MHz timings look spectacular. Memtweakit 40000 is great. I get 36500. Haswells are all different. You may have one of them that has to trade CPU OC for higher memory clocks. I've heard about that, but not seen a comparison like you did here.

Have fun

Jeff

jfrech14
Level 7
How did you get that???!! My computer won't even boot at 1.54 volts. Is there a setting that overrides it? When I go to 1.54 it just goes to bios and says overvolt fails. It was quite frustrating haha. Thanks again for all the help. I may give it another try sometime soon. Would getting another 8 gb of ram help or does it test efficiency not capacity? I feel like getting another video card or upgrading would be the best upgrade but not now. I'm gonna try to get the 4.6 with those tight timings and see if I can't overvolt more. I got a mem tweakit score of 80601. I found out that my xmp changes the refresh from 2000 to 5000 so I brought it back to 2200 with tight timings and the score jumped a ton
Asus Maximus VI Extreme
Intel i7 4770K
8 GB G.Skill Trident X 2600MHz
1TB Samsung Evo SSD
1TB Western Digital Caviar Black
Corsair HX1000W PSU
SLI - Custom Watercooled EVGA GTX 770 SC 4GB
XSPC Custom Watercooling loop AX360 dual pump with 360 Rad and 240 Rad
NZXT Switch 810
NZXT Hue LED lighting

jab383
Level 13
I've never run into a BIOS overvolt limit because I don't boot at high voltage. My BIOS profile with all the benchmark optimization has the cores at 4700MHz, 100.0MHz Bclk and 1.500Vcore. That boots and is Realbench stable. Then I tweak in the extra voltage and Bclk in Windows.

Perhaps AI Suite can do this, but I don't use AI Suite for several reasons -- it's not even installed. Way down the directory on the Asus distribution disk is a handy utility called TurboV Core. It's a no-nonsense OC tweaker that runs in Windows. It doesn't try to help you by setting any limits, so you have to be careful. That's the voice of experience. I made some typo -- wrong value, slipped decimal point, wrong box, something -- and toasted a very good CPU. With TurboV, and I think with any tweaking software running in the operating system, take small steps when reducing voltage. You never can tell what will happen when voltage gets so low that it causes errors.

TurboV Core works on a nice subset of clocks and voltages and is very handy for quickly stepping voltages up or down when seeking a low power stable OC. In this case, I use it to get Vcore up after boot, then increase Bclk.

Jeff

jfrech14
Level 7
I tried to tweak it in ai suite 3 but no luck. I just took off the lid and am doing bare die cooling. Non overclocked 3900 idle is at 27C with about a 20-24 C ambient but I suspect the difference only shines at high temperatures so I am going to test. I ended up only being able to squeeze out a 108000 on mem tweakit but reloaded my os before I uploaded it like an idiot so my official score is 107555 or something. I started to get memory not installed errors so I chickened out haha. I really want to try and tweak ithe CPU but I'm worried I won't do it correctly and fry my CPU so I'm going to play it cool (ha cool.. Pun) for a little while and just plan my next attack. I nicked the green coating when removing the heat sink which I'm thankful didn't do anything because I can't see a tiny but of copper wiring and it had me ****ting bricks until I saw the benchmark at 88000 haha. But even a safe clock of 4.4 gets me at 80000 about so I think I'm in the 70 percentile in the lottery and can't complain after what I put it through. It's fun learning from you guys. Scary but fun 🙂
Asus Maximus VI Extreme
Intel i7 4770K
8 GB G.Skill Trident X 2600MHz
1TB Samsung Evo SSD
1TB Western Digital Caviar Black
Corsair HX1000W PSU
SLI - Custom Watercooled EVGA GTX 770 SC 4GB
XSPC Custom Watercooling loop AX360 dual pump with 360 Rad and 240 Rad
NZXT Switch 810
NZXT Hue LED lighting

Zka17
Level 16
27C idle temp with 20-24C ambient is very-very good with Haswell! How your max core temps changed with delidding?

jfrech14
Level 7
I am getting data for overclocks and what not. Using IntelBurnTest on Maximum (Which uses between 6000-6100MB of memory every time at 2600 MHz @ 1.65600 Volts I am getting data using the logging feature in RealTemp and getting the voltages using ROGConnect. The first point was all stock configurations for CPU and XMP for RAM. Second was 4.2 GHZ, then 4.4 and I am doing 4.6 now so I will post the results tonight. It definitely is a nonlinear gain in temps/volt which is probably why I am so unstable at 4.7 so I will see about it. Because I still can't get my system to not BSOD past 1.54Volts even when tuning them using ROG Connect in the OS nor AISuite. 4.4 GHz @ 1.262 Volts Maxed out on Core1 at 68C with an average of 60C and a mode of 66C which could have been lower if I opened up the top vent where the triple rad is intaking air and opened up the case but I'm trying to get real life situations here so I'll will just call that fortuitous forgetfulness because I always overclock with my rig open. The tests I am running are bare die which I expected more cooling than I saw but I recall before delidding I didn't use realtemp so the temperature comparison can't be scientifically founded but I am guessing between 15 and 25C drop in temperatures. probably up to 30C on core3 which has shown to stay the coolest. I am frustrated I can't get past 1.54 volts but I have a TON on my SSD now and dont really want to delete it all just to test fresh. I may use my WD Black to test but I don't think performance will be as good. My SSD clocked a performance in turbo mode of 1GB/s which dumbfounded me. I may try to save up for a 4930K and in a year or two try a Rampage series test rig but that's far off down the road. I have been really impressed with those rampage boards and at the time of building this rig it was too much for me but rethinking it I may have gone with it but I love my system. Wish I went with a corsair 900D though since the cabling under the hood is horrendous in the NZXT not to mention I could get better temps with a dual loop which I will migrate to later. Depends on the next intel chip. I have never passed 40C on either of my GPUs and they have been through the gambit except for Crysis 3 of course haha. Actually 35C is high for these cards so I can only imagine the stability of the CPU and GPUs under separate loops. Considered putting my H100i on the cpu to see if a closed loop on only the cpu would help but since it is bare die I want to touch the waterblock as rarely as often so I won't grind down the silicon. I am worried about contact though. I have it fairly tight. I loosened them some and tightened them to where medium pressure finger tight is all they are and have small thermal pads surround it to help out to not bend the board. I also applied only the thinnest amount of thermal paste because I think I had been puttting it on rather thick so I put it on and spread it out so that you could nearly see though it but it was fully covered and made sure where the die touches the block was flat and minimum pitting. So.. hopefully all is well. I think mid 50s for 4.2 100% load is decent so I am not too worried about that. I just hope they put the voltage regulators off the cpu next time around. Better to have a little more head on mobo than built heat up on chip. However I don't think the tock version can take them off can they since the 87 chipset boards don't have them.. unless they kept them on there in case intel was like woah.. bad idea. Let's change that haha. Well... anyway.. I am letting my liquid temps lower and settle before running another test so I will post results tonight hopefully. Cheers!
Josh
Asus Maximus VI Extreme
Intel i7 4770K
8 GB G.Skill Trident X 2600MHz
1TB Samsung Evo SSD
1TB Western Digital Caviar Black
Corsair HX1000W PSU
SLI - Custom Watercooled EVGA GTX 770 SC 4GB
XSPC Custom Watercooling loop AX360 dual pump with 360 Rad and 240 Rad
NZXT Switch 810
NZXT Hue LED lighting

jfrech14
Level 7
Doing this intel burn test has shown me some weird stuff. At 4.6 ghz @ 1.46v it ran 20 out of 30 of my tests on max reaching 93c on core 1 and 83 on the others and just randomly stopped. Temps were fine. Just not enough volts I guess. Even though it already ran 20 successfully.. Odd to me. Also using rog connect showed me setting my voltage at 1.519 volts doesn't meat crap because it's reading a vcore of 1.55... Um what? Just going on record to say intel was pretty stupid when they did the whole throttling thing. I have officially reached tj max and witnessed the throttling.. They throttled the frequency.. THE HEAT IS FROM THE VOLTAGE! Haha who cares about a bsod from overclocking failure. They need to throttle vcore not frequency. If it's supposed to do that.. Mine didn't haha. It kept pumping 1.55 volts during ibt but throttled it to 4.4. Bad idea to me. Just saying. And 4.4 ghz @ 1.262 didn't pass realbench but it was 10c cooler than ibt. So realbench requires more voltage than ibt but realbench at the same voltage as ibt is cooler. I guess those even out once realbench is stable at a higher voltage
Asus Maximus VI Extreme
Intel i7 4770K
8 GB G.Skill Trident X 2600MHz
1TB Samsung Evo SSD
1TB Western Digital Caviar Black
Corsair HX1000W PSU
SLI - Custom Watercooled EVGA GTX 770 SC 4GB
XSPC Custom Watercooling loop AX360 dual pump with 360 Rad and 240 Rad
NZXT Switch 810
NZXT Hue LED lighting